The following paragraphs include a general discussion of lithography, an explanation of latent imagery, and some discussion of the methods (including latent image measurements) that lithographers use to determine whether their processes are working properly.
Many modern semiconductor fabrication processes involve the deposition of a photosensitive resist material upon a substrate such as a wafer (which may have various material layers formed upon it). The resist material is then exposed to radiation of a particular frequency (or to particles) through a reticle. The radiation interacts with the resist material and produces a pattern which may be considered a three-dimensional distribution of chemical species within the resist. This three-dimensional distribution within the resist is termed a "latent image." Generally speaking, there is desirably a strong correlation between any horizontal (i.e., parallel to the plane of the substrate or wafer) cross-section through the resist material and the image (as spatially filtered by a lens) that the reticle was designed to generate.
In typical semiconductor processing, the resist material containing these latent images may be processed through a number of subsequent steps, and either the exposed or unexposed portion of the resist material is then removed using either wet or dry techniques. (Whether the exposed or unexposed material is removed depends on whether the resist is a "positive" or "negative" resist.) Typical subsequent semiconductor processing often involves steps, such as etching, ion implantation, or chemical modification of the substrate material from which the photoresist has been removed.
The faithfulness of the three-dimensional resist feature (i.e., the latent image formed by the processes mentioned above) in replicating what is desired by the process engineer is controlled by a large number of variables. It is the task of the lithographer to maximize this faithfulness by judicious equipment selection, and adjustment, and definition of usable process windows.
The metrics used by those versed in the art for characterizing the faithfulness of resist feature formation have been the subject of a number of journal articles. In the article, Mack et al., "Understanding Focus Effects In Submicron Optical Lithography, Part 2, Photoresist Defects," SPIE Vol. 1088 (1989), the authors discuss the interaction of the aerial image (i.e., the three-dimensional image in free space) with the photoresist and model the interaction mathematically to determine what features of the aerial image are important in determining lithographic performance. In addition, the publication: Mack, "Photoresist Process Optimization," KTI Microelectronics Seminar Interface 87 (1987), pp. 153-167, discusses various factors which affect the shape of the latent image.
The majority of current practices regarding characterization of faithfulness of the lithographic process to that desired by the process engineer falls into three categories:
i) quantitative measurement of linewidth in etched layers under varying conditions of exposure and focus, PA0 ii) semiquantitative characterization by manual observation (using optical observation or some form of SEM) of features formed under varying conditions, PA0 iii) direct interrogation of the aerial image by scanned slits or image dissecting artifacts. PA0 Imax is the maximum intensity of the scattered signal from the i.sup.th alignment mark in the X or Y dimension PA0 Imin is the minimum intensity of the scattered signal from the i.sup.th alignment mark in the X or Y dimension.
Some of these will be discussed in more detail.
Practitioners have consistently sought various methods for monitoring the accuracy of the process of transferring the reticle pattern into raised or recessed features on the substrate (i.e., a wafer or various material layers upon it). Since the process involves many steps, it is desirable to analyze and understand each step individually. For example, it is axiomatic that the latent image be as faithful a reproduction of the intended portions of the reticle pattern as possible in order to eventually produce acceptable semiconductor features such as lines, spaces, etc. To evaluate the faithfulness, in the past, practitioners have frequently created latent image lines in photoresist, developed these lines and then etched the line patterns into the underlying substrates of test wafers. Then these test wafers were sectioned and subjected to SEM analysis. The linewidths in the SEM were then measured and compared with the desired linewidth. Unfortunately this monitoring process introduces many additional variables and possible sources of inaccuracy. For example, the developing process, the etching process, and the SEM measurement process may all contribute to inaccuracies in the comparison between the final etched line and the desired reticle "standard." (Experienced practitioners will also realize that the above procedure may be complicated by etching proximity effects.)
Another technique used to evaluate the adequacy of lithographic processes is to expose the photoresist (to radiation projected from a reticle pattern) thus creating a latent image, and then develop the photoresist. The developed photoresist will exhibit a series of raised features which may be examined and compared with the intended design pattern on the reticle. For example, the developed features may be evaluated by SEM, typically using either high accelerating voltages and conductive coatings on the features or low accelerating voltages coupled with tilted substrate surfaces. The developed features may also be observed and measured optically.
Two factors (characteristic parameters) which are among those that are very important in lithographic processes (and are normally adjusted on a routine basis) are the focus and exposure dose ("fluence") of the stepper or other tool used to produce latent (and then developed) images. Either poor focus or poor exposure dose will create latent images with poorly defined edge profiles and, therefore, ultimately will cause poorly defined edge profiles or unacceptable width variations on etched features or implanted regions. (For example, an exposure dose that is too low provides insufficient energy to induce adequate chemical change in the resist.) Good focus and exposure are often determined experimentally by the following procedure. Either focus or exposure dose is held constant while the other is varied during a series of exposures in which sets of lines and spaces are created in the photoresist. Measurements of the linewidth deviation from nominal of either the developed photoresist image or an etched feature are graphically plotted with the variable focus or exposure dose to determine, for example, best focus for a constant exposure dose. The graphical technique often produces curves which, because of their contours, are sometimes called "smile plots."
Recently, one stepper manufacturer has utilized latent imagery in an attempt to determine proper stepper focus. The procedure is as follows: A flat, resist-coated wafer is put in the stepper. A row of approximately 40 local alignment mark features is sequentially exposed in a straight line proceeding horizontally across the entire wafer. The exposure dose is maintained constant while focus is varied for each alignment mark location. Thus a sequence of latent images is created in the photoresist. Each alignment mark consists of several groups of horizontal and vertical lines which are rectangular in shape and which are normally used to provide an "X" and a "Y" (or "horizontal" and "vertical") alignment reading. The stepper is then programmed to move the wafer containing the latent images of these alignment marks to a position under its local alignment optics. An alignment is performed at each location and the signal contrast for the X and Y alignments is measured and stored for each alignment mark location. The data is analyzed in an effort to determine proper focus.
It may be noted that in the normal operation of the stepper's local alignment system the scattered light signal from the edges of a developed or etched alignment mark is measured as the mark is scanned in X and Y coordinates under the system's illuminating reference beams. (These beams are arranged in shape, spacing and orientation so that they will match the edges of alignment marks.) (The term "scattered light" is used here to indicate light which returns along a path different from that followed by reflected or transmitted light. Because the optics are arranged to observe only the scattered light, the system performs as a dark field microscope and is termed a "dark field alignment system.") However, in the practice described above, the procedure is performed upon a latent image.
After an image scattering signal for the X and Y (i.e., horizontal and vertical) alignment of each alignment mark is measured, a "contrast" number for the X and Y alignment of each alignment mark is determined. The "contrast" is given for the X and Y alignment of each alignment mark as ##EQU1## where K.sub.i,X,Y is the contrast for the i.sup.th alignment mark for X or Y, i=1 . . . 40
The X and Y data (i.e., K.sub.iX and K.sub.iY for each alignment mark on the wafer) are combined (i.e., averaged) to provide an average contrast for each alignment mark. The average contrast for each alignment mark is plotted against mark X-Y location (i.e., the focus used for each mark exposure). The maximum for the latent image alignment contrast curve thus obtained is used as a reference focus to be chosen as "machine focus" for the stepper. "Machine focus" is an average reference focus which will be discussed in further detail below.